Hi Carlos, Charlie, and All,
Apologize for the confusions caused! Here is what I am doing:
0. to compute the relative solvation free energy difference between two small organic molecules, e.g. L0, L1
1. setup TI with PMEMD where the dual topology is used (i.e. both L0 and L1 are included in one single prmtop file)
2. run the TI simulations using PMEMD with "icfe" and "ifsc" turned on; "timask1" and "timask2" include L0 and L1, respectively; "scmask1" and "scmask2" include the unique groups in L0 and L1, respectively.The ele and vdw are switched from L0 to L1 simultaneously.
3. Using trajectories generated in step (2), the individual frames were extracted into the .inpcrd files using PARMED. Based on the individual coordinate files, (1) change vdw parameters of L0 or L1, (2) regenerate the prmtop file (i.e. the new Hamiltonian), (3) recompute the energies for each of the coordinate files using the new prmtop file and the same setting of TI to get the new energies.
4. Using the reweighting protocol with the energies from both the original (from which the trajectories were generated) and new Hamiltonian (where non-bonded parameters were changed) to reoptimize the force field parameters to better reproduce the relative HFE between L0 and L1.
From what I am doing now, step 3 is too time consuming. I am wondering if there is an alternative way to do this energy rerun more efficiently? I tried imin=5 with sander.MPI. However, I cannot have sander.MPI read the PMEMD generated trajectories with TI.
Any suggestion and comment is appreciated!
Best,
Mingjun
> -----Original Messages-----
> From: "Carlos Simmerling" <carlos.simmerling.gmail.com>
> Sent Time: 2019-10-18 01:44:04 (Friday)
> To: "AMBER Mailing List" <amber.ambermd.org>
> Cc:
> Subject: Re: [AMBER] more efficient TI rerun over existing trajectory
>
> can you tell us how you are actually doing it? for example, do you use
> imin=5 in sander, or are you reading each frame as a separate sander/pmemd
> run? we need to know the details in order to suggest improvements.
>
> On Thu, Oct 17, 2019 at 9:35 AM 杨明俊 <mjyang.dicp.ac.cn> wrote:
>
> > Hi Carlos,
> >
> > Thanks for your response. What I am working on is trying to improve the
> > production of hydration free energy by adjusting atomic charges and vdw
> > parameters. To this end, I firstly run the simulation with the original
> > Hamiltonian, e.g. H0. Then I changed the selected atomic charges or vdw
> > parameters, resulting in a new Hamiltonian, e.g. H1. I'd like to know if
> > the adjustment of charge or vdw parameters can produce better predicted HFE
> > against experimental measurements. To reduce the computational cost, I used
> > reweighting method to get the HFE under H1 over the trajectory (T0)
> > generated with H0. Thus I need to rerun the energy calculations using H1
> > over T0. The problem is when too many frames are recorded in T0, it will
> > take too much time to finish the rerun, especially when MBAR or WHAM is
> > used (which requires Nframe x Nstate x Nstate energy calculations). Thus
> > I'd like to know if there is a way to do the energy rerun more efficiently
> > with TI.
> >
> > Thanks for any suggestions and comments!
> >
> > Mingjun
> >
> >
> > > -----Original Messages-----
> > > From: "Carlos Simmerling" <carlos.simmerling.gmail.com>
> > > Sent Time: 2019-10-17 18:54:45 (Thursday)
> > > To: "AMBER Mailing List" <amber.ambermd.org>
> > > Cc:
> > > Subject: Re: [AMBER] more efficient TI rerun over existing trajectory
> > >
> > > Can you tell us exactly what you are doing, so we can see if it might be
> > > improved? Also what do you mean by "very painful" - which aspect
> > > specifically would you like to improve?
> > >
> > > On Wed, Oct 16, 2019, 11:25 PM 杨明俊 <mjyang.dicp.ac.cn> wrote:
> > >
> > > > Dear Everyone,
> > > >
> > > >
> > > > I am studying a reweighting scheme with TI for relative hydration
> > free
> > > > energy calculation, for which I need to rerun the trajectory with a
> > > > modified Hamiltonian. However, I haven't figured out an efficient way
> > to
> > > > rerun the trajectory. Currently, I firstly extract individual frames
> > from
> > > > the trajectory and then recompute the energy for this frame with
> > different
> > > > Hamiltonian. However, this is very painful when there are too many
> > frames
> > > > included in the trajectory. Does anyone know a way to do the rerun
> > with TI
> > > > over existing trajectory more efficiently?
> > > >
> > > >
> > > > Thanks in advance!
> > > >
> > > >
> > > > Mingjun
> > > > _______________________________________________
> > > > AMBER mailing list
> > > > AMBER.ambermd.org
> > > > http://lists.ambermd.org/mailman/listinfo/amber
> > > >
> > > _______________________________________________
> > > AMBER mailing list
> > > AMBER.ambermd.org
> > > http://lists.ambermd.org/mailman/listinfo/amber
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Oct 28 2019 - 06:30:03 PDT
